The current project used a synthesis of social and developmental psychological theories to develop an intervention to promote science motivation and ethnic/racial and gender identity among American Indian middle school adolescents. Results suggested that the intervention was not successful with one exception: Adolescents who were matched with a role model on both ethnicity and gender increased in performance approach goal orientations from pre-intervention to one-week post-intervention. In the following sections, I briefly review the results of the study and offer potential explanations regarding why most hypotheses were not supported. To
conclude, I discuss limitations of the study and present directions for future research in this area. The Effect of the Self-Affirmation Intervention
The primary purpose of the intervention was to increase students’ science motivation—in particular, their science self-efficacy, science interest, and mastery and performance goal
orientations in science via reading a biography about a scientist and completing of a self- affirmation task. In Hypothesis 1, I predicted that adolescents who engaged in the self-
affirmation task would report increased science motivation. Neither adolescents who engaged nor those who did not engage in the self-affirmation task increased in science motivation either at intervention or one-week post-intervention. Thus, Hypothesis 1 was not supported.
Other empirical evidence regarding self-affirmation interventions is relatively mixed in success of improving academic outcomes in adolescents. Cohen et al. (2006) employed a multi- session values affirmation intervention with African American and European American seventh
grade students. Students were randomly assigned to an affirmation or control condition. Those who were assigned to the affirmation condition engaged in three to five affirmation writing exercises (affirming the self positively) over the course of an academic year. African American students, but not European American students, who engaged in the affirmation condition had significantly improved their GPAs compared to African Americans in the control conditions (Cohen et al., 2006). Longitudinal effects of the intervention were observed three years post- intervention: Affirmed African Americans continued to improve their GPAs over their non- affirmed African American counterparts (Cohen et al., 2009). European American students showed no benefits of the self-affirmation. Sherman et al. (2013) conducted a study that was methodologically similar to that of Cohen et al. (2006) with White and Hispanic/Latino American sixth, seventh, and eighth grade students. Results revealed that Hispanic/Latino American students who engaged in the affirmation writing had higher GPAs than their non- affirmed Hispanic/Latino American counterparts by the end of the first year. Effects of the intervention endured two years post-intervention, as affirmed Hispanic/Latino American students continued to earn higher GPAs than non-affirmed Hispanic/Latino American students (Sherman et al., 2013). However, once again, European Americans did not benefit from the intervention, and other researchers who have aimed to replicate the intervention benefits found by Cohen et al., (2006) and Sherman et al., (2013) with other adolescent samples and have had
mixed/marginal success (Borman, Grigg, & Hanselman, 2016; Bratter, Rowley, & Chukray, 2016; de Jong, Jellesma, Koomen, & de Jong, 2016; Dee, 2016; Hanselman, Rozek, Grigg, & Borman, 2016).
Three key methodological differences between the current study and prior successful affirmation interventions with adolescents could explain why no effect of intervention was
observed in the current study. The first two differences are the frequency of affirmations and the timing of testing for intervention effects. Of published affirmation interventions that were successful in improving academic outcomes for adolescents, each have had at least two
affirmation sessions (usually toward the beginning of the academic year) for participants in the experimental groups, and effects of intervention were observed over the course of an academic year, typically by grading quarter or semester (Bowen et al., 2013; Bratter et al., 2016; Cohen et al., 2006; Dee, 2016; Hanselman et al., 2016; Sherman et al., 2013). In the current study,
adolescents had only one self-affirmation session, and the effects of intervention were tested immediately after the intervention and one-week post-intervention. Thus, results from the current study provide support for the notion that self-affirmations tasks, in insolation, are likely not beneficial for promoting academic motivation in the short term. Further, results are not inconsistent with the possibility that the effect of self-affirmations may be subtle and gradual, unfolding over time (especially in conjunction with continued sessions of self-affirmations).
The third difference that might explain why no effect of intervention was observed in the current study could be the choice of intervention outcomes (i.e., science self-efficacy, science individual interest, and mastery and performance approach goal orientations in science). To date, much of the education intervention research employing self-affirmation tasks has focused on academic achievement as the outcome of interest. It is possible that no effect was observed in the intervention because of selection of motivational measures. Self-efficacy, individual interest, and mastery and performance approach goal orientations measure motivational beliefs. However, perhaps the intervention was not effective in changing these beliefs, but may have led to change in other facets of motivational beliefs or behavior. For example, students who engaged in the self-affirmation condition, reading about the effort employed by a scientist to reach his or her
professional goals, might have an increased belief in the importance of effort. Such students could be more motivated in the future to put effort into their science homework and tests. Future research aiming to increase motivation via self-affirmation studies could consider using a broad array of motivation measures to assess what forms or kinds of motivation such interventions may affect.
Taken together, it is likely that these three methodological differences; (1) single
frequency self-affirmation, (2) immediate tests of intervention effects, and (3) different measures of outcome variables jointly contributed to explain why no effect of intervention was observed. It should be noted that self-affirmation intervention studies are sensitive to subtle methodological and contextual changes, as these changes may disrupt the targeted underlying psychological processes (Yeager & Walton, 2011). Thus, perhaps the changes outlined above were too divergent from previously successful self-affirmation studies. Future research would benefit from better understanding how the efficacy and fidelity of self-affirmation interventions may be affected by various methodological differences in study designs and contextual differences in the environment.
The Effect of Role Model Matching on Self-Affirmation Efficacy
For Hypotheses 2 and 3, more nuanced analyses of the intervention efficacy were
examined to understand the potential effect of reading about a scientist who matched the students on both ethnicity and gender, only gender, only ethnicity, or on neither ethnicity nor gender. Results provided little support for Hypotheses 2 and 3, as none of the outcome variables changed, on average, among students in any of the experimental conditions. Ad hoc probing showed increases in performance goal orientation among students who were matched to their
role model on both ethnicity and gender, whereas students in the other conditions reported decreased mastery motivation.
In general, results suggest that reading a brief text about a successful scientist who had the same gender and ethnicity as the student was insufficient to change students’ science
motivation and their ethnic or gender identity. Based on role model theory and research (Bryant & Zimmerman, 2003; Lockwood & Kunda, 1997; Marx & Ko, 2012; Zirkel, 2002), I had expected that matching the role model to the student on each of the social identities would independently enhance treatment effects. However, the current study differs from much of the role model literature, as students in this study were assigned a role model. In most of the prior role model research, students are asked to list role models in their lives. After a list is generated, more descriptive demographic information is typically gathered about the role model to ensure that a role model who matches the youth on ethnicity and/or gender is listed (Bryant &
Zimmerman, 2003; Hurd et al., 2009; Yancey et al., 2002; Zirkel, 2002). Perhaps it was an unrealistic assumption that students would view the scientist they had read about in the intervention as a role model. A manipulation check assessing perceived similarity and attainability could have been added to the study to ensure that students viewed the scientists featured in the biography as role models. Indeed, this practice was used in an empirical study that employed the use of a “novel” role model assigned to participants in an adult sample (Marx & Ko, 2012).
Another possible explanation for lack of findings may stem from the length and content of the biographies. In general, the biographies were short, only containing three paragraphs of approximately five sentences per paragraph. Even the youngest students in the sample (i.e. 6th graders) were able to read through the biographies in less than 10 minutes. In these biographies,
students read about how the scientists came to be interested in science, the degrees they earned, the scientists’ current employment, and their science-related service. It is possible that the information in the biographies was not relevant enough for the students to be able to identify or associate themselves with the scientist. Increasing the quantity of content in the biographies and ensuring that content was personally relevant to the adolescent might create a stronger role model effect, especially use of information that highlighted the ethnic and gender identities of the scientists. An example of creating more general relevance would be to explain that the scientists went to the same middle school as the students. An example of creating more specific relevance to the students on the basis of a social identity would be to explain a time when the scientist was in middle school and someone told the scientist that s/he could not do well in science because the student was a girl or American Indian (or both). However, despite the negative feedback, the scientist persisted and earned their degrees in science.
One significant intervention effect was observed for students who were matched to the scientist on ethnicity and gender: Those students reported increased performance goals in science after the intervention. In contrast, mean performance goals of other groups decreased after
treatment. This result may be indicative that matched role models are an effective mechanism for change in specific kinds of outcomes. Out of four science motivation variables, performance approach goal orientation could be considered a form of extrinsic motivation, whereas the other three variables (self-efficacy, interest, and mastery approach goal orientations) are forms of intrinsic motivation. Individuals who endorse performance approach goal orientations in school aim to demonstrate their competence through academic achievement as shown through exam and course grades. Other studies examining the effects of matched role models have found that students who reported matched role models were more likely to have higher GPAs (Bryant &
Zimmerman, 2003, Zirkel, 2002). Role model theory suggests that an individual who identifies another person as a role model is likely already intrinsically motivated/interested in the domain in which that role model excels. However, the skills, milestones, and accomplishments that the role model has already achieved is what differentiates the individual from the role model (Morgenroth et al., 2015). This difference between the role model and individual could act as a mechanism for increases in extrinsic motivation, as the individuals seeks to emulate the role model by achieving the same skills, milestones, and accomplishments as the role model. Thus, perhaps role models and role model matching may be particularly effective in boosting more extrinsic motivation, like performance approach goal orientations or general achievement. Indeed, empirical research from the role model literature often uses an individual’s performance or achievement as a measure of role model effectiveness (e.g., Hoyt, Burnette, & Innella, 2012; Latu, Schmid Mast, Lammers, & Bombari, 2013, O’Brien, Hitti, Shaffer, Van Camp, Henry, & Gilbert, 2016).
Alternatively, this significant result could be spurious. Many analyses were conducted in assessment of the intervention efficacy; thus, at least one significant effect might to emerge due to chance. Further, examination of group means from pre-intervention to one-week post-
intervention showed that group means were regressing to an overall sample mean (i.e.,
performance approach goal orientation means at pre-intervention and post-intervention for those who were matched to the scientist were Ms = 4.21 and 4.62, respectively, and those in all other experimental conditions and the control condition were Ms = 4.81 and 4.54, respectively). Thus, the interpretation of this result must be cautious and not overstated.
Centrality as a Predictor of Intervention of Efficacy
In Hypothesis 4 I expected that ethnic and gender centrality would significantly predict the efficacy of the intervention, such that the students with higher levels of centrality in the respective centrality-match conditions would report greater increases in science motivation, compared to students with lower levels of centrality. However, ethnic and gender centrality were not significant predictors of intervention efficacy and results did not support Hypothesis 4. Perhaps the most obvious factor in explaining why ethnic and gender centrality were not significant predictors is that there was no effect of treatment for any of the science motivation measures. With very little observed change from pre-intervention to intervention and post-
intervention in science motivation scores, it is not surprising that ethnic and gender centrality did not significantly predict intervention efficacy.
However, even if a treatment effect had been observed in the experimental conditions, it could be the case that ethnic centrality (as well as other dimensions of ethnic identity) are too rudimentary during early adolescence to have moderated treatment effects. Ethnic identity development theory suggests that nuanced dimensions of ethnic identity such as ethnic centrality and ethnic private and public regard are only beginning to develop in late childhood and early adolescence (Umaña-Taylor et al., 2014). Indeed, a study of individual differences in racial centrality of African American youth showed that the stability of racial centrality increased from early to middle adolescence (Hoffman, Kurtz-Costes, Rowley, & Adams, 2017). Thus, given its nascent form, perhaps it is not surprising that centrality did not significantly predict treatment efficacy in the intervention.
Limitations and Directions for Future Research
The study had several limitations that might inform future interventions to boost students’ motivation and social identities that are congruent with academic success. First, there appeared to be a novelty and/or experimenter effect in the mean scores of the key study variables across the three sessions of the study. Relatively consistently, pre-intervention scores across all groups were the highest of the three sessions. Motivation scores at intervention and post-intervention were typically lower than pre-intervention scores and were more similar to each other across conditions. It is likely that many of the students had never participated in a psychology research study before participating in the current study, and seeing some of these measures for the first time might have resulted in higher endorsement. Further, especially at the first session, students may have anticipated that the experimenter would want the students to have higher endorsement of the various measures. To circumvent this issue in future studies, it would be helpful to offer two or three pre-intervention sessions across several weeks. Scores across the pre-intervention sessions could then be averaged to create a more accurate measure of baseline mean levels of various measures.
An assumption underlying the study design was that the scientists featured in the
biographies would be considered role models for the students. However, much of the role model research with children and adolescents requires participants to identify their role models, who are therefore personally known to them. This method ensures that an individual can serve as a role model for the participant. However, in the current study, students were assigned a scientist role model and it is not known whether or not these scientists were viewed as role models for the students. Certainly, for those students who were interested in and who enjoy science, it is likely that these scientists would serve as a role model. However, for students who were not interested
in or enjoyed science, it is not clear if and how the scientists could serve as role models for these students. Thus, in future role model research where role models are assigned to the participant, it would be important to use a manipulation check to ensure that role models serve their purpose. As previously mentioned, scholars have identified the perceived relevance of the role model to the individual and the believed attainability of the role model’s success as important factors (Lockwood & Kunda, 1997). Therefore, in future research, study designs where participants are matched to a role model should include questions that ask the extent to which participants believe that role model is relevant to them and the extent to which the role model’s success is attainable.
Finally, it is important to consider that the sample represents a specific population, even within the larger American Indian and Alaskan Native population. It is not clear if and how results would be different among other Native Nations or ethnic/racial minority groups in the United States. To date, there are 567 federally recognized Native Nations in the United States. Of these Nations, only 46 Nations, or approximately 8%, are located east of the Mississippi River. The Nation represented in this study is one of those 46 federally recognized Native Nations east of the Mississippi River. That Nation has a unique history compared to many of the Native Nations in the western parts of the United States. Histories for these Nations in particular include forced assimilation (including attendance at Indian boarding schools) and The Indian Removal Act of 1830. Negative effects of these historical traumas have been documented (Brave Heart, 2003) and are known to influence the content and development of ethnic identity of American Indian youth (Brown, Dickerson, & D’Amico, 2016). Thus, the historical trauma experienced by this Nation likely shapes its citizens’ ethnic identity, views about education, general worldviews, and perhaps, even their response to the current intervention. With such ties to historical trauma,
perhaps young citizens are skeptical of academia and higher education as a means for a future career and lifestyle. Thus, students from the current sample may be less responsive to the intervention than students from other Nations in the western United States. Conversely, because of the forced assimilation and loss of Native culture due to assimilation, young citizens from this Nation may be more likely to view academia and higher education as means for success in their future and thus may be more responsive to the intervention compared to students from other Nations.
Despite the limitations listed above, results of the current study provide valuable information regarding the methodological design and implementation of self-affirmation intervention studies. Results suggest that the fidelity of self-affirmation interventions are relatively fragile and sensitive to methodological and contextual changes. Future research will benefit from a more nuanced understanding of which methodological factors and contextual